Abstract:According to the disadvantages of artificial immune algorithm to search for long time and low efficiency, it was made some improvement, so as to keep the population diversity, at the same time improve the convergence speed.To reduce the trial and error time of sheet metal forming process parameters, numerical simulation was used to establish the approximate model.To box as an example, software Dynaform was used to obtain the training datas to establish the RBF neural network approximation model.RBF neural network was optimized by the artificial immune algorithm to get the position and the number of hidden layer centrals, and the output layer weights were determined by the pseudo inverse method.The model was optimized by using artificial immune algorithm improved to obtain the load curve of variable blank holding force. Research results show that the optimized variable pressure curve can effectively improve the quality of sheet metal forming.